Re: [Hampshire] Linux RAM usages

[Chris Dennis]

On 13/07/10 20:22, Sean Gibbins wrote:

On 13/07/10 17:59, James Courtier-Dutton wrote:

Hi,

I notice that Linux tends to always use about 95% of RAM all the time,
for cache mostly.
Is there any sysctl that could move this to 80% ?
Please don't ask me why because the answer it stupid!!!

Kind Regards

James


Did anyone else initially read the subject as 'Linux RAM sausages'?

*sigh*

Sean



No.  You must be hungry.

Chris
--
Chris Dennis  cgden...@btinternet.com
Fordingbridge, Hampshire, UK

--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

Re: [Hampshire] Linux RAM usages

[James]

On Tue, 2010-07-13 at 20:23 +0100, Dee Earley wrote:
> 
> That sounds more like an IO/bandwidth problem to me.
> 

Well, the raw throughput is always very good... 

James

-- 
James   theholyet...@googlemail.com
PGP key ID: 03F94B5D
---


signature.asc
Description: This is a digitally signed message part
--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

Re: [Hampshire] Linux RAM usages

[Dee Earley]

On 13/07/2010 19:30, James wrote:

On Tue, 2010-07-13 at 19:01 +0100, Hugo Mills wrote:

Not that I'm aware of. Linux uses all the spare RAM for cache,
because having spare RAM unused would be a waste. If a process
actually needs more RAM (because it's starting up and allocating
memory, say), then some files are evicted from cache to make space.


It uses it *very* aggressively, though, and seemingly with no way to
restrain it. To the point where copying a few large (order gigabytes)
files will cause the cache to grow to over 3/4 physical, and push out a
lot of stuff allocated to applications I'm using (including X). And then
interactivity just plummets. (I seem to remember a certain scientific
app that uses large amounts of RAM getting swapped out, too.)


That sounds more like an IO/bandwidth problem to me.

--
Dee Earley (d...@earlsoft.co.uk)

irc:irc://irc.blitzed.org/
web:http://www.earlsoft.co.uk
phone:  +44 (0)780 8369596

--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

Re: [Hampshire] Linux RAM usages

[Sean Gibbins]

On 13/07/10 17:59, James Courtier-Dutton wrote:

Hi,

I notice that Linux tends to always use about 95% of RAM all the time,
for cache mostly.
Is there any sysctl that could move this to 80% ?
Please don't ask me why because the answer it stupid!!!

Kind Regards

James
   


Did anyone else initially read the subject as 'Linux RAM sausages'?

*sigh*

Sean

--
music, film, comics, books, rants and drivel:

www.funkygibbins.me.uk


--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

Re: [Hampshire] Linux RAM usages

[Vic]

>I've never seen this happen -- even when copying large files
> around.

What do you have set for /proc/sys/vm/swappiness?

> I would suggest that you have something else going on.

Firefox is the one that usually sets my machine off. Eventually, it gets
to a GC loop that doesn't finish :-(

Vic.


--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

Re: [Hampshire] Linux RAM usages

[Hugo Mills]

On Tue, Jul 13, 2010 at 07:30:04PM +0100, James wrote:
> On Tue, 2010-07-13 at 19:01 +0100, Hugo Mills wrote:
> >Not that I'm aware of. Linux uses all the spare RAM for cache,
> > because having spare RAM unused would be a waste. If a process
> > actually needs more RAM (because it's starting up and allocating
> > memory, say), then some files are evicted from cache to make space.
> 
> It uses it *very* aggressively, though, and seemingly with no way to
> restrain it. To the point where copying a few large (order gigabytes)
> files will cause the cache to grow to over 3/4 physical, and push out a
> lot of stuff allocated to applications I'm using (including X). And then
> interactivity just plummets. (I seem to remember a certain scientific
> app that uses large amounts of RAM getting swapped out, too.)

   I've never seen this happen -- even when copying large files
around. I would suggest that you have something else going on.

   Hugo.

-- 
=== Hugo Mills: h...@... carfax.org.uk | darksatanic.net | lug.org.uk ===
  PGP key: 515C238D from wwwkeys.eu.pgp.net or http://www.carfax.org.uk
  --- Your problem is that you've got too much taste to be ---   
a web developer. 


signature.asc
Description: Digital signature
--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

Re: [Hampshire] Linux RAM usages

[James]

On Tue, 2010-07-13 at 19:01 +0100, Hugo Mills wrote:
>Not that I'm aware of. Linux uses all the spare RAM for cache,
> because having spare RAM unused would be a waste. If a process
> actually needs more RAM (because it's starting up and allocating
> memory, say), then some files are evicted from cache to make space.

It uses it *very* aggressively, though, and seemingly with no way to
restrain it. To the point where copying a few large (order gigabytes)
files will cause the cache to grow to over 3/4 physical, and push out a
lot of stuff allocated to applications I'm using (including X). And then
interactivity just plummets. (I seem to remember a certain scientific
app that uses large amounts of RAM getting swapped out, too.)

Maybe there's some better setting for the writeback timeouts I could
try.

>There's a "swappiness" control that governs how much working set
> (i.e. the memory actually used by applications and not otherwise on
> disk) is swapped out to the pagefile, and how aggressively, but that
> won't help you with the filesystem cache bits.

I remember trying to tune swappiness down hard in the recent past and
arrived at the conclusion that the knob did pretty much nothing.
Certainly did not reduce Linux's tendency to swap out memory allocated
by processes in favour of the cache.

James

-- 
James   theholyet...@googlemail.com
PGP key ID: 03F94B5D
---


signature.asc
Description: This is a digitally signed message part
--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

Re: [Hampshire] Linux RAM usages

[Hugo Mills]

On Tue, Jul 13, 2010 at 05:59:46PM +0100, James Courtier-Dutton wrote:
> I notice that Linux tends to always use about 95% of RAM all the time,
> for cache mostly.
> Is there any sysctl that could move this to 80% ?

   Not that I'm aware of. Linux uses all the spare RAM for cache,
because having spare RAM unused would be a waste. If a process
actually needs more RAM (because it's starting up and allocating
memory, say), then some files are evicted from cache to make space.

   There's a "swappiness" control that governs how much working set
(i.e. the memory actually used by applications and not otherwise on
disk) is swapped out to the pagefile, and how aggressively, but that
won't help you with the filesystem cache bits.

> Please don't ask me why because the answer it stupid!!!

   Why? :)

   (If the answer is, "someone thinks that the machine is short of RAM
and they need more", then some education is in order).

   Hugo.

-- 
=== Hugo Mills: h...@... carfax.org.uk | darksatanic.net | lug.org.uk ===
  PGP key: 515C238D from wwwkeys.eu.pgp.net or http://www.carfax.org.uk
--- SCSI is usually fixed by remembering that it needs three --- 
terminations: One at each end of the chain. And the goat.
 


signature.asc
Description: Digital signature
--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

Re: [Hampshire] [OT, I suppose] Router

[Simon Reap]

Roger Munford wrote:
Why not disconnect the domestic circuit from the main BT box and then 
connect the router to see if the fault persists. If it doesn't the 
fault lies in your domestic circuit so you can methodically work your 
way through the circuit connecting parts until the fault re-occurs. If 
the fault persists with nothing but the router connected and you are 
confident that the router is OK then it probably is a BT fault.


That's what BT Broadband would get you to do before they sent out an 
engineer.  If your main phone socket has an unscrewable fromt plate, 
undo that and plug the router directly into the socket under the plate 
(that will disconnect your domestic wiring).


Simon



--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

[Hampshire] Linux RAM usages

[James Courtier-Dutton]

Hi,

I notice that Linux tends to always use about 95% of RAM all the time,
for cache mostly.
Is there any sysctl that could move this to 80% ?
Please don't ask me why because the answer it stupid!!!

Kind Regards

James

--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

Re: [Hampshire] [OT, I suppose] Router

[Vic]

> Why not disconnect the domestic circuit from the main BT box and then
> connect the router to see if the fault persists.

This is part of the standard test before most ISPs will even talk to BT.

You have to find the master socket, unscrew the front plate, and pull it
off. Behind that, there is another socket - the "test socket". If the
router connects when plugged into the test socket, but not when the rest
of the wiring is connected, BT will charge your ISP for the callout (and
they will charge you).

Vic.


--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

Re: [Hampshire] [OT, I suppose] Router

[Tim]

It is standard practice for BT to charge is they don't find a fault or the 
fault 
is on the users wiring\equipment.

You could try phoning BT and telling them that you have a problem with the 
phone 
line, I had to do this with a certain ISP as they were adamant the fault I was 
reporting was my router, turned out somebody had cut the cable (builder doing 
work for the business next door).

Tim

On Tuesday 13 July 2010 16:14:58 Owain Clarke wrote:
> 
> 
> 
>   
> 
> 
> I'm still working on this one (the main symptom being random outages),
> and have been unable to trace what the problem is.  My ISP (the Phone
> Coop) tell me that if they call in an Open Reach engineer and it turns
> out that the problem is within my property I'll be liable for a hefty
> BT callout fee, so I wonder if anyone knows of someone who would test a
> domestic network and charge a reasonable amount of money?
> 
> Owain
> 
> cite="mid:aanlktind7qhxgzgxfcqrvjjp5l9hqvv_qevec8bwk...@mail.gmail.com"
>  type="cite">
> On 8 July 2010 18:34, Owain Clarke   dir="ltr"><  href="mailto:simb...@cooptel.net";>simb...@cooptel.net>
> wrote:
>   style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt
> 0.8ex; padding-left: 1ex;">After a period of connection problems my ISP
> wants to test my line with a different router, and I wonder if there's
> anyone who might be able to lend me one for a couple of weeks. I'm in the
> Romsey area, but work also takes me near Fareham, Brockenhurst and New
> Milton
>   
> Thanks in anticipation
>   
> Owain
> 
> 
>   
> 
> 
> 
> 



--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

Re: [Hampshire] [OT, I suppose] Router

[Roger Munford]

Why not disconnect the domestic circuit from the main BT box and then 
connect the router to see if the fault persists. If it doesn't the fault 
lies in your domestic circuit so you can methodically work your way 
through the circuit connecting parts until the fault re-occurs. If the 
fault persists with nothing but the router connected and you are 
confident that the router is OK then it probably is a BT fault.


Roger

On 13/07/10 16:14, Owain Clarke wrote:
I'm still working on this one (the main symptom being random outages), 
and have been unable to trace what the problem is.  My ISP (the Phone 
Coop) tell me that if they call in an Open Reach engineer and it turns 
out that the problem is within my property I'll be liable for a hefty 
BT callout fee, so I wonder if anyone knows of someone who would test 
a domestic network and charge a reasonable amount of money?


Owain
On 8 July 2010 18:34, Owain Clarke > wrote:


After a period of connection problems my ISP wants to test my
line with a different router, and I wonder if there's anyone who
might be able to lend me one for a couple of weeks. I'm in the
Romsey area, but work also takes me near Fareham, Brockenhurst
and New Milton

Thanks in anticipation

Owain




--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--
--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--

Re: [Hampshire] [OT, I suppose] Router

[Owain Clarke]

I'm still working on this one (the main symptom being random outages),
and have been unable to trace what the problem is.  My ISP (the Phone
Coop) tell me that if they call in an Open Reach engineer and it turns
out that the problem is within my property I'll be liable for a hefty
BT callout fee, so I wonder if anyone knows of someone who would test a
domestic network and charge a reasonable amount of money?

Owain

  
On 8 July 2010 18:34, Owain Clarke 
wrote:
After
a period of connection problems my ISP wants to test my line with a
different router, and I wonder if there's anyone who might be able to
lend me one for a couple of weeks. I'm in the Romsey area, but work
also takes me near Fareham, Brockenhurst and New Milton
  
Thanks in anticipation
  
Owain


  





--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--